Skip to main content

Table 2 The list of operations in the segmentation and feature computation stages and the sources of the CPU and GPU versions

From: Scalable analysis of Big pathology image data cohorts using efficient methods and high-performance computing strategies

Feature Computations

Class

Operations

Computed Features

CPU and GPU Implementation

Pixel Statistics

Histogram Calculation

Mean, Median, Min, Max, 25 %, 50 %, and 75 % quartile

Implemented

Gradient Statistics

Gradient and Histogram Calculation

Mean, Median, Min, Max, 25 %, 50 %, and 75 % quartile

Implemented

Haralick

Normalization pixel values and Co-occurrence matrix

Inertia, Energy, Entropy, Homogeneity, Max prob, Cluster shade, Prominence

Implemented

Edge

Canny and Sobel

Canny area, Sobel area

OpenCV (Canny), Implemented (Sobel)

Morphometry

Pixel counting, Dist. among points, Area and Perimeter, Fitting ellipse, Bounding box, Convex hull, Connected components, Area, Perimeter, Equivalent diameter, Compactness, Major/Minor axis length, Orientation, Eccentricity, Aspect ratio, Convex area, Euler number

Implemented

  1. We used the implementation of the morphological reconstruction operation by Vincent for the implementation of several segmentation operations. Implemented indicates our implementation of the respective operations